AI systems that fail predictably, not creatively.
AI/ML: inference runtimes · agent harnesses & sandboxed execution · guardrails & evals (LLM-as-judge, golden datasets) · human-in-the-loop design · replayable, deterministic pipelines
Systems: on-prem platform buildouts · LLMOps · distributed infrastructure · agent observability
Previously: high-scale ad marketplaces (sub-ms p99, 600k+ RPS) · trading infrastructure at D. E. Shaw · scientific computing at NASA
Speaker — PyCon India 2020: "Python Serialization and the Hallows of Parallel Computing" — yes, pickle is still cursed
Publication — Multi-Stage Tracking for Precise Localization of Targets in Video Sequences · Expert Systems with Applications (Elsevier)
Education — MS Computer Science, NYU · B.Tech Mathematics & Computing, DTU
Reach — rsaim.info · LinkedIn · Medium · saimorsaim [at] gmail.com
Languages
AI Systems
Infrastructure




